xingchen

This MCP server wraps iFlytek workflows and exposes them as MCP tools, enabling LLM-driven orchestration of iFlytek workflows within your MCP-enabled environment. It supports sequential, parallel, looping, and nested execution through the MCP routing and node system, with streaming output via a hook mechanism for real-time feedback. You can prepare a config.yaml that describes your workflows and authenticate against the iFlytek platform to fetch workflow metadata and credentials as needed. The server acts as a bridge: you define a workflow in the cloud, configure an MCP client to load and execute that workflow, and the MCP engine handles data passing, variable I/O, and model orchestration across nodes and tools.

Prerequisites:

• Python and uvx installed on your machine or environment where MCPs run
• Git installed and network access to fetch repositories
• Access to the iFlytek workflow hub if needed

Step-by-step installation:

1. Install uvx (Python/uv):

   • Ensure Python 3.8+ is installed
   • Install uvx if it is not already available in your environment (e.g., via pipx or your preferred method)

2. Install the MCP server package from GitHub:

   • Run the following command to fetch and install the server payload for MCP usage:

     uvx --from git+https://github.com/iflytek/ifly-workflow-mcp-server ifly_workflow_mcp_server

3. Prepare configuration:

   • Create a config.yaml (or use mcp.json) that defines your flow(s) and authentication as required by the server.

   • Example path usage in the MCP config:

     CONFIG_PATH=/path/to/your/config.yaml

4. Run the MCP server through uvx using the configured server name (as defined in mcp_config):

   • Example (from your MCP client): uvx --from git+https://github.com/iflytek/ifly-workflow-mcp-server ifly_workflow_mcp_server

5. Verify operation:

   • Use your MCP client to list, instantiate, and trigger workflows via the server.
   • Check logs for authentication, network access, and node execution details.

Tips and common notes:

• Ensure CONFIG_PATH (or equivalent) points to a valid YAML/JSON configuration describing your workflows and authentication credentials.
• If you encounter connectivity issues, verify network access to the iFlytek workflow endpoints and that your API keys are valid.
• When using streaming outputs, enable the hook mechanism in your workflow to receive real-time results.
• For local development, you can point CONFIG_PATH to a sample config to validate basic MCP routing before connecting to the cloud.
• This server supports multiple node types and MoM-based model selection; you can leverage these to try different model combinations at critical workflow stages.